Download How Important is Genetics for an Understanding of Evolution?1

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Genetic engineering wikipedia , lookup

Public health genomics wikipedia , lookup

Biology and consumer behaviour wikipedia , lookup

Twin study wikipedia , lookup

Genome evolution wikipedia , lookup

Epistasis wikipedia , lookup

Gene expression programming wikipedia , lookup

Gene wikipedia , lookup

Adaptive evolution in the human genome wikipedia , lookup

Behavioural genetics wikipedia , lookup

Dual inheritance theory wikipedia , lookup

Medical genetics wikipedia , lookup

Designer baby wikipedia , lookup

History of genetic engineering wikipedia , lookup

Genome (book) wikipedia , lookup

Genetic drift wikipedia , lookup

Heritability of IQ wikipedia , lookup

Quantitative trait locus wikipedia , lookup

Group selection wikipedia , lookup

Koinophilia wikipedia , lookup

Human genetic variation wikipedia , lookup

Polymorphism (biology) wikipedia , lookup

Population genetics wikipedia , lookup

Microevolution wikipedia , lookup

Transcript
AMER. ZOOL., 26:811-820 (1986)
How Important is Genetics for an Understanding of Evolution?1
R. C. LEWONTIN
Museum of Comparative Zoology, Harvard University,
Cambridge, Massachusetts 02138
SYNOPSIS. The contributions of modern genetics to the understanding of evolution have
been threefold. First, it has documented the extent of genetic variation that exists in
populations as a basis for future evolution. In particular it has shown that natural selection
has not destroyed all variation, as might be predicted, so that there must be mechanisms
for the maintenance and origin of new selectable variation that balance the homogenizing
forces of selection. Second, it has begun to provide a mechanism of the origin of genetic
novelties which must be at the basis of the major features of the history of life. In doing
so, it has shown how selection for new features is at all times historically contingent and
that evolution is at all times at risk of falling into genetic dead-ends. Third, modern
genetics has greatly enriched the diversity of mechanisms known to cause evolutionary
change. All of these mechanisms involve the conversion of variation between individuals
into variation between populations in time and space, but many are non-selective or even
counter-selective. Natural selection is not the only mechanism of evolution.
gage of any student of biology? The answer
is, "Not much," if we restrict ourselves to
the domain of evolutionary theory at its
simplest level.
We all know, for example, that Mendelism saved Darwin's theory of evolution by
natural selection from what appeared to
be a fatal contradiction. Darwin's theory
amounts to claiming that the differences
that appear between populations and
species in space and time are the concentrations of differences that already exist
between individual organisms within populations. Natural selection (as well as other,
non-selective population processes) causes
an increase in the frequency of some heritable types in some populations as comof the Origin of the Species (1937) and Mayr's pared with other populations and so the
Animal Species and Evolution (1963). Evo- populations diverge in their collective
lution is "descent with modification" so the properties. Such a theory depends absorules of descent are obviously at the basis lutely, then, on the existence of inter-indiof the rules of evolution. Indeed, we might vidual variation as a condition of evolution.
go as far as Dobzhansky (1951) and say that But theories of continuous blending inher"evolution is a change in the genetic com- itance predict that sexual reproduction will
position of populations" and so conflate result in a homogenization of a population
the study of evolution with the study of and a rapid loss of intra-population variapopulation genetics, pure and simple. Is tion. In fact, the variance in a phenotypic
there really any more to be said than is character will decrease by a factor of Vz in
already packed into the intellectual bag- each succeeding generation of blending.
Thus, Darwin's theory, dependent as it is
on standing variation, could not be right
1
in a world of blending inheritance. MenFrom the Symposium on Science as a Way of Knowing—Genetics presented at the Annual Meeting of the del's principle of segregation, as embodied
American Society of Zoologists, 27-30 December in the Hardy-Weinberg equilibrium, its
INTRODUCTION
At first sight it seems absurd to ask about
the contribution of genetics to our understanding of the evolutionary process. After
all, it is obvious that neo-Darwinism is precisely the union of concepts from Mendelian genetics with the Darwinian theory of
natural selection. The dependence of evolutionary explanations on basic principles
of genetics is manifested by the demand
that students study genetics before evolution, by the initial chapters on population
genetics in textbooks of evolution (some of
which turn out to be little more than an
explanation of population genetic concepts), and by the great classics of evolutionary synthesis like Dobzhansky's Genetics
1985, at Baltimore, Maryland.
811
812
R. C. LEWONTIN
mathematical expression, guaranteed the
maintenance of heritable variation despite
the union of gametes from different parents.
Unfortunately, although Mendelism
saved Darwinism from sex, it could not save
natural selection from itself. The problem
of the modulation of variation by reproduction quite aside, evolution by natural
selection is a self-negating and self-defeating process. Beginning with a population
that has somehow acquired heritable variations in traits relevant to survival and
reproduction, natural selection will cause
an enrichment of those heritable phenotypes, "the fit," with the highest reproductive rates until the population consists
only of these most fit types. That is, beginning with a heterogeneous assemblage of
organisms, selection produces a homogeneous assemblage among whom there is
nothing left to select. "Selection" on the
one hand demands variants among which
to select, and on the other hand disposes
of variants, destroying the very fuel that
drives the motor of evolution. If evolution
is indeed "a change in gene frequencies of
populations," then genetics must somehow
show how continued change in gene frequencies has been possible over the last 2
billion years. And it is precisely this task
that has been addressed by molecular,
developmental and population genetics and
which has been the great contribution of
the science of genetics to evolutionary biology. That is, genetics has saved naive Darwinism, with its exclusive emphasis on
selection driving populations to a fixed most
fit type, from foundering on the rocks of
its own contradictory formulation. It has
done so by three paths:
1) Population and biometrical genetics
has shown that, for whatever reason, there
is considerable heritable variation in virtually any population, on which selection
may act. That is, selection has not, in fact,
exhausted variation.
2) Molecular and development genetics
have elucidated how new heritable variation can arise, and can provide the basis
for genetical novelties beyond the limits of
variation that already exists, into the indefinite future.
3) Developmental and population genetics have vastly enriched our knowledge
of the repertoire of evolutionary forces,
beyond simple selection, to explain the
diversity and continued diversification of
organisms.
THE EXISTENCE OF DIVERSITY
The most direct evidence of the existence of heritable variation on which selection can operate has come from artificial
selection experiments. The great mass of
such experiments has been carried out in
agriculturally important animals and plants,
but these might be regarded as atypical in
several respects. First, a rather limited
range of traits has been selected, usually
those directly concerned with yield of seed,
milk, flesh, etc. that are of direct agronomic importance. Yet even these often
involve complex physiological and anatomical changes that amount to remaking the
organism. The selection for machine harvestable tomatoes required changing the
growth habit of the plant, replacing indeterminate by determinate growth, making
both flowering and fruit ripening virtually
synchronous, changing the shape and texture of the fruit and the length of the growing season, among other traits. A more
serious limitation on the generality of
results from agronomic selection programs
has been the previous breeding history of
the stocks on which selection has been carried out. The population for selection has
been sometimes derived from a closed flock
or herd, in which case the genetic base is
very restricted, sometimes from an historical variety in which case the original
genetic variation is unknown, but most
often from a deliberate cross between
divergent populations in order to maximize the variation for selection. In any case,
the relation to natural populations is
doubtful. The outcome of such selection
experiments is that almost always it is possible to make some progress, showing that
there is heritable variation for a remarkable variety of phenotypic traits. It is the
exceptions, however, that are most revealing for a general understanding of evolution. For example, there has been no real
progress in breaking the "one egg a day"
GENETICS IN EVOLUTION
barrier in chickens, although there is plenty
of genetic variation for egg production in
poultry. Progress in selecting heat resistant
varieties of cattle has also had indifferent
success. Apparently, the complex physiological interactions that limit egg production to a daily rhythm, or that come
together to provide high heat tolerance to
productive cattle cannot easily be reorganized by mass selection directly on the trait
itself. Some of this kind of limitation on
natural selection comes from the very large
number of so-far unbreakable genetic correlations between physiological traits in
agronomically important organisms. So, for
example, yield is negatively correlated with
protein content in soybeans while nicotine
content and carcinogenic tars are positively correlated in tobacco. These
unbreakable correlations may represent
purely historically contingent genetic correlations that might have turned out differently had the past history of selection
and breeding been different, or they may
be developmental and biochemical correlations that could not be broken by any
constellation of genes that would still produce functioning organisms. In either case,
they demonstrate the extremely important
general evolutionary principle that despite
heritable variation in separate characters
in a population, selection may be unable to
drive a response to some particular combination of character states. In this way
correlated characters may resist selection
and preserve variation that can be the subject of other selective forces at other times.
Artificial selection experiments on laboratory organisms derived from natural
populations, chiefly various species of Drosophila, have shown unambiguously that
extraordinary amounts of genetic variation for physiological and anatomical traits
exist in nature. It is a commonplace of Drosophila geneticists that anything is selectable, from body size (Robertson and Reeves,
1952) through pattern of body parts (for
example, Maynard Smith and Sondhi,
1960), to the amount of recombination in
particular chromosome regions (Chinnici,
1971). From the mass of such experiments,
the student of evolution is forced to conclude that heritable variation is present and
813
selectable for a remarkable variety of anatomical and physiological properties. But
not for all. Attempts to select for anatomical asymmetry in Drosophila, for example,
have repeatedly failed to produce strains
that are on the average larger on the leftor right-hand sides, despite the fact that
individual flies are all slightly asymmetrical, but with no average right- or left-hand
bias in the population as a whole. It is, however, possible to select for fluctuating asymmetry, that is for a population of flies each
of whom is individually more asymmetrical, but again with no population bias
toward right or left sides (Mather, 1953).
This last case, is an example of an extremely
important general result of selection
experiments that is nevertheless misunderstood even by many geneticists. An
organism is not determined by its genes,
nor even by its genes and the environment
in which it develops acting jointly, but by
genes, environment, and random developmental factors acting at the level of single cell divisions and thermal noise. The
left- and right-hand sides of a developing
Drosophila are genetically identical, and, in
any conventional meaning of the word
"environment," have experienced the same
sequence of developmental environments.
Yet the left- and right-hand side of a fly
will have different bristle numbers, different eye facet numbers and slightly different
wing lengths. These differences have arisen
by "developmental noise" during cell division and cell differentiation. There are then
three distinct levels at which phenotypic
variation occurs. First, in a fixed environment, or on the average over some specified distribution of environments, groups
of organisms may differ in the average value
of a trait because the groups differ genetically. Second, even if organisms have the
same phenotype when exposed to a given
developmental environment, they may differ in another environment. Moreover,
heritable differences between organisms in
one environment may be completely
reversed in other environments. (See, for
example, Gupta and Lewontin, 1982.)
Finally, there is variation in the stability of
developing organisms to microscopic perturbations in development during cell divi-
814
R. C. LEWONTIN
sions and the growth and development of
single cells and patches of tissue. What is
significant for the understanding of evolutionary processes is that all three levels
may be influenced by different sets of genes
and may be subject to natural selection separately. That is, we may select to increase
say, the average eye size of Drosophila at a
given temperature; we can independently
select eye size to have a greater or smaller
response to developmental temperature
(Waddington, 1960) and we can also select
for lesser or greater individual fluctuating
asymmetry at a given temperature (Mather,
1953). Natural selection can act on the
average expression of a trait, on the
responsiveness of the trait to changing
environments, and on the internal developmental stability of a trait.
We must continue to bear in mind that
most of the traits successfully selected for
have been developmentally homogeneous
and restricted, as for example eye facet
number or the number of bristles on a particular region of the cuticle. Thus, a few
genes or even a single segregating locus
may contribute most of the genetic variation. More developmentally complex or
generalized characters are not easily
selected and may, as previously discussed,
have unbreakable genetic correlations or
else may not be under genetic control at
all. So, although it is possible to select,
independently, for greater and lesser
developmental stability of eye size and wing
variation in Drosophila, there are no genes
for general developmental stability that
effect both systems simultaneously (Scharloo, 1964; Lewontin, unpublished).
One of the most puzzling features of the
observed diversity of organisms is the frequent contrast between the phenotypic
homogeneity within groups and the diversity between groups. A good deal of the
practice of systematics depends on the ability to make absolute distinctions between
species, genes, or families, based on characters that are invariant within the group.
Thus, a diagnostic character for the family
Drosophilidae, as opposed to other families
of Acalyptratae, is the presence of one proclinate and two reclinate orbital bristles.
Every individual of every species of the
family possesses this character, unlike all
individuals of every species of, say, Dacus
(the med fly) lacks it. In a theory of evolution that depends upon variation within
populations for the source of variation
between species, how do we account for
these phenotypically constant traits? Are
those traits never again to be the basis of
future evolution, since they are now without selectable variation? One possibility, of
course, is that there was once genetic variation for these characters, but that this
variation was lost during the process of
species differentiation and has not yet been
recovered within species. Undoubtedly
such purely historical explanations must be
correct sometimes. The present environments of the species may select against
intraspecific variations, whereas at a former time mutational variation was tolerated, and so, accumulated. Such hypotheses are, alas, unverifiable in particular cases.
There is, however, another explanation for
phenotypic uniformity, which has been
uncovered by selection experiments, and
which provides a very different insight
into the possibilities of future evolution.
This is the phenomenon of canalization.
All individuals of all species of Drosophila
have two large anterior and two large posterior bristles on the scutellum. Using a
mutation to disrupt the development of
bristles, flies can be produced that have 0,
1 or 2 bristles instead of four. The average
number of bristles in the mutant flies can
be increased by selection so that 3 and then
4 bristle flies will appear. At the same time,
among non-mutant sibs of the 4 bristle flies,
some flies with 5 bristles will appear. There
is now phenotypic variation in scutellar
bristle number among non-mutant flies and
this variation can be made use of to select
6, 7, 8 . . . bristle flies. So, apparently all
the while there was selectable genotypic variation for bristle number, when there was
phenotypic uniformity (see Rendel [1967] for
his summary of these and similar experiments). The phenomenon of developmental constancy, despite genotypic (or
environmental) variation, called canalization by Waddington (1942), is itself the consequence of genes for developmental buffering. These genes themselves are varying
GENETICS IN EVOLUTION
in populations and can be selected so as to
increase or decrease the intensity of canalization (Rendel, 1967). Developmental
buffering is not unlimited in its range. If a
sufficiently strong perturbation of development occurs, either because a new
mutant gene has drastically interfered with
the developmental process, or because a
major shift in environment has occurred,
then the genetic variation for the trait will
become manifest as phenotypic variation
and natural selection may operate. Evolution of an apparently constant trait may
then occur episodically as major environmental or genetic shifts occur, revealing
the underlying genetic variation that was
all the while present.
While selection experiments have
revealed a fund of heritable variation present in natural populations, the description
of that variation in genetic terms has not
been possible because of the complex relations between phenotype and genotype. We
cannot know from such experiments, for
example, whether there are many segregating genes or only a few and whether
there are many alternate alleles at these
loci, or at what frequencies they are.
Answers to these questions are important
to the evolutionist because they provide
knowledge of the long term prospect for
selection, as opposed to the immediate rate
of response revealed in selection experiments. Some information about the variation of single genes in populations has
existed for a long time, chiefly genes whose
products could be detected immunologically, as for example the blood groups in
humans and cattle. Unfortunately, only
genes that were already known to be variable could be studied in this way, so that
no one can say how many red cell antigens
are, in fact, invariant in populations. The
gene-by-gene study of variation in natural
populations was greatly enhanced by the
introduction of protein gel electrophoresis
into evolutionary studies (Harris, 1966;
Hubby and Lewontin, 1966). This technique has made it possible, over the last 20
years, to study scores of gene loci (more
than 100 in humans) in hundreds of species.
Not all classes of genes are equally represented in these studies which have, for
815
purely technical reasons, concentrated on
soluble proteins, mostly enzymes, while
finding out much less about membrane
proteins, particle-bound enzymes and other
classes of insoluble polypeptides. Unfortunately, no satisfactory method yet exists
for a study of these latter, nor do we know
what fraction of the genome they represent. When gel-electrophoresis was first
introduced it was thought that only a fraction, perhaps about one-third, of the amino
acid variation in proteins could be detected,
since electrophoresis depends upon charge
differences in proteins to discriminate
them. Advances in technique (Singh et al.,
1976) and reconstruction experiments in
which proteins of known amino acid composition and three-dimensional structure
were subject to electrophoresis (Ramshaw
et al., 1979) have shown that their early
doubts were exaggerated. For the most
part, genes shown to be monomorphic or
polymorphic with only two or three alleles
segregating in populations, by earlier cruder methods, have proven to be accurately
described. A few genes with larger numbers of segregating alleles like xanthene,
dehydrogenase and esterase in Drosophila, are
now known to be vastly more polymorphic
than originally thought, with as many as
33 alleles segregating in a single population (Keith, 1983; Keith et al., 1985), but
these appear to be an unusually highly
polymorphic class. As a consequence of the
vast effort of evolutionary geneticists
studying genie variation in plants, animals,
vertebrates, invertebrates, prokaryotes and
eukaryotes, it is now possible to make some
strong generalizations about genie variation in natural populations. About !/s of all
loci (based largely on loci coding for soluble proteins) are polymorphic in a typical
species while % have essentially only a single allele characterizing the entire species.
Of course, all loci have occasional mutations, but these rare events are not counted
as making a locus polymorphic. Characterized in another way, in a typical diploid
individual about 10% of its genome is heterozygous, while another 20% is homozygous for alleles that are varying within the
species but happen to be homozygous in
the individual. Thus, a typical Drosophila
816
R. C. LEWONTIN
species would be polymorphic for, say,
3,000 loci, and typical individual fly heterozygous at 1,000 loci, if there are on the
order of 10,000 genes coding for soluble
proteins.
If the various alleles at each of these
polymorphic loci were sufficiently different
in their biochemical properties and activity
to be reflected in the physiology and development of the whole organism, then we
would be certain that an immense fund of
selectable genetic variation was, in fact,
present in nearly all species and that for
some reason, the previous history of selection had not resulted in a genetically homogeneous collection as might be expected
from simple theory. But we cannot be sure
that all this genetic diversity is indeed physiologically important. It may turn out that
many of the variant alleles in natural populations are functionally equivalent, at least
at the level of the integrated physiology of
the whole organism, so that species are
indeed rather homogeneous from the
standpoint of natural selection. That does
not mean that the species could not evolve,
but that molecular evolution would consist
largely of the accidental and random
replacement of one molecular form with a
functionally equivalent one. Thus the
molecular description would evolve, but not
the function. It may indeed be that many
of the amino acid replacements that have
occurred in the evolution of molecules like
cytochrome-c or haemoglobin over
hundreds of millions of years are precisely
of this nature. We appear then to be in a
quandary. The purpose of studying molecular variation in natural populations was
to determine whether, in fact, there is
enough genetic variation for natural selection to continue to act whenever circumstances of life change, or whether selection
exhausts relevant variation so that adaptive
evolution is stalled for long periods for lack
of something to select. The study has shown
that a lot of genetic variation exists, but,
in itself, the observation of that variation
is insufficient to determine its possible significance to natural selection.
Most attempts to resolve this problem in
the past have concentrated on detecting
the action of natural selection in one of
two ways. The first has been to make theoretical predictions about the statistics of
genetic variation: how many alleles, their
frequencies, the differences in their frequencies from population to population.
These models are based both on selective
and non-selective theories of genetic variation, in the hope that the actual data would
discriminate between the models. These
attempts have failed. Given the known processes of mutation, migration, fluctuations
in population size, and varying kinds of
natural selection, there are simply too many
undetermined parameters to allow the theories to be rigorously tested. The second
approach has been to try to measure
directly the physiological or fitness differences between different genotypes. While
some biochemical and physiological differences have indeed been detected in a few
cases, of which alcohol dehydrogenase in
Drosophila is the best documented (see
Lewontin [1985] for a review), it has not
in general been possible to measure the
selective differences since, if they exist, they
will be small.
A quite different way to resolve the problem appears if we consider the monomorphic
rather than the polymorphic loci, or, what
is equivalent, if we look within the allelic
classes of the polymorphic genes. If a locus
is monomorphic because selection is
enforcing a homogeneity of amino acid
sequence for functional reasons, then alleles
that are identical in kind will not necessarily
be identical in ancestry, at least in the recent
past. On the other hand, if the monomorphism is purely a historical accident of the
loss of new, functionally equivalent mutations through the inbreeding that occurs
in finite populations, then we should expect
all the alleles identical by kind also to be
identical by ancestry. They will all be
descended from a recent common ancestral gene in a common ancestor. But how
can we distinguish between identity by kind
and identity by historical descent if the
genes are identical? By looking at their
DNA sequences. Because the genetic code
is degenerate there are multiple codons,
usually differing in their third positions,
that specify the same amino acid. If a locus
is homozygous by selective constraint we
GENETICS IN EVOLUTION
should expect to see mutational variation
in third positions (and in introns) despite
homogeneity in amino acid sequence. If,
on the other hand, the monomorphism is
the result of recent common ancestry, then
the third positions and introns will also be
monomorphic in the population. The first
such comparison of silent DNA polymorphism with amino acid polymorphism has
been made for the alcohol dehydrogenase
817
elty. While this has not been a major theme
of evolutionary genetics, a number of very
illuminating discoveries have been made
by molecular geneticists that mark out the
territory for what must eventually become
a major preoccupation of genetics and evolution.
A major advance in our understanding
of the acquisition of new function has come
from the study of the human globin gene
gene in Drosophila melanogaster by Kreit- family. Adult human hemoglobin consists
man (1983). In a sample of genomes taken almost entirely of a tetramer of two alpha
from a geographical range including chains and two beta chains. These four
Africa, Europe, Japan, and North Amer- chains with their four haem groups form
ica, Kreitman found a 7% polymorphism a structurally flexible unit that changes its
of DNA bases in introns and third posi- shape during oxygenation and makes postions, but absolutely no variation in amino sible the cooperative effect that allows each
acids except for the single major two-allele added oxygen to attach more easily than
electrophoretic polymorphism already the one before. The genes that code for
known. Within the electrophoretic alleles the alpha and beta chains are on separate
there was complete uniformity of amino chromosomes, but when the amino acid
acids but a high level of silent base varia- sequences of the chains are compared they
tion. Since %U of all random DNA base are essentially the same length (141 amino
changes should cause an amino acid change, acids in alpha and 146 in beta) and are
yet none were found, the evidence for con- identical in 62 amino acids, more similar
sistent selection against amino acid substi- than human beta hemoglobin is to that of
tutions is overwhelming. The monomor- the shark. It is clear that the differentiation
phism is a selective not a historical one. of alpha and beta chains arose by a dupliThis makes it somewhat more likely that cation of genes, probably around the time
the single amino acid polymorphism that of origin of the bony fishes. This duplicais observed is maintained by some sort of tion made possible the heterodimeric
balance of forces, but the case is not proved hemoglobin molecule with its greatly
and we might equally hold that it is the increased efficiency of function. But that
single variant that is tolerated by an oth- is only the beginning of the story. In addierwise very discriminating selection. The tion to the alpha and beta chains, four other
ability to distinguish historical from selec- globin chains, delta, gamma, epsilon, and
tive identity is a unique feature of studies zeta appear and disappear during develof DNA variation and for this reason it is opment. The delta, gamma, and epsilon
studies at this level which represent the chains differ from beta by only 10, 39 and
future of experimental population genetics. 36 amino acids, respectively. Moreover, the
genes for epsilon, gamma, delta and beta
have very similar exon-intron structure and
THE ORIGIN OF NOVELTY
are arranged along chromosome 11 in the
Evolution cannot simply be the replace- order within an extremely short distance
ment of one variant allele at a locus by of each other (about 50 kilobases). Clearly,
another. The major features of evolution this subfamily of genes has been derived
involve the acquisition of new repertoires one from another by gene duplication with
of biochemical, anatomical and behavioral some divergence. In like manner the alpha
traits that are often added to, rather than subfamily consists of genes all in a 40 kithe simple replacement of alternative lobase region of chromosome 16. What is
forms. A satisfactory Darwinism must remarkable about those two subfamilies of
therefore be more than the population genes is that their arrangement along the
genetics of shuffling old variation. It must chromosome exactly parallels their
include the phenomenon of genetic nov-
818
R. C. LEWONTIN
appearance and disappearance in development. Embryos begin life with zeta and
epsilon as the alpha-like beta-like components of their hemoglobin. Those then give
way to alpha and gamma, making foetal
hemoglobin, which then become, postnatally alpha and beta with small amount
of delta. Presumably the four different
hemoglobin tetramers that are produced
at different development stages correspond to physiological exigencies of each
stage. We see then that a differentiation
and multiplication of function can arise by
gene duplication followed by gene differentiation. That this differentiation sometimes goes awry is shown by the presence
within both the alpha and beta gene clusters of pseudogenes, DN A sequences almost
identical with the other of the family, but
with fatal stop-code mutations that prevent
them from coding proteins.
Another source of genetic novelty may
be the conversion of a gene of redundant
function to a totally new activity. Such a
conversion has been demonstrated experimentally by Hall (1978) using the ebg gene
in E. coli. This gene is induced by lactose
but its enzyme has a low activity on lactose
so its function in wild-type E. coli is unclear,
but definitely expendable. Hall attempted
to select mutants of this gene that would
ferment a new substrate, lactobionate. He
succeeded, but the order of events was
extremely instructive. First, it was necessary to obtain a constitutive mutant, since
lactose would not be available as an inducer. This regulatory mutation, a separate
event, is a necessary first step. Second, direct
selection with lactobionate produced nothing. It was necessary to build up the new
function in three successive selections. First
selection with lactose produced lactose fermenters. Some of these (Class II) were also
fermenters of a related sugar lactulose while
others were not (Class I). When Class I and
Class II were selected with lactobionate
again there was no success. When original
strains were directly selected with lactulose, a positive strain (Class II) was produced, but this could not be further evolved
to lactobionate fermentation. The only
evolutionary pathway that succeeded was
first selection with lactose to produce Class
I, followed by selection with lactulose to
produce Class IV, followed by selection with
lactobionate to produce the final result. All
other pathways and shortcuts were deadends. We see then that four separate mutational events (including the regulatory
mutation) must succeed each other in the
correct order and that other orders
although they produce phenotypically intermediate stages in the evolution, are genetically cul-de-sacs. Evolving a new function is
like threading a maze and temporary forward progress may be at the expense of
eventual success. Evolution is an historically contingent process which does not
allow any form to arise from any other.
The structure of accessibility in evolution
reflects in part the history of particular
mutations that have occurred historically.
What Hall's experiment shows is that evolution consists not only of opening new possibilities but of closing off others at the
same time.
ALTERNATIVES TO ADAPTIVE SELECTION
The earliest contribution of genetics to
evolutionary theory was to show, both theoretically and experimentally that natural
selection of small variations in phenotype
could explain continued evolution, as Darwin claimed. This was, in fact, the burden
of Fisher's Genetical Theory of Natural Selec-
tion (1930) and Haldane's Causes of Evolution (1931). The irony is that the development of population genetics since that
time has shown that while direct natural
selection of character states may be a sufficient explanation of evolution, it is not a
necessary one and that many other forces
and phenomena are causes of evolution.
First, random historical forces may operate independently of or even contrary to
selection. As shown by Wright (1932), in
a finite population alleles with no selective
advantage will be fixed by random drift
causing random differentiation in space and
time. Even deleterious genes can be fixed
provided the product of N, the effective
population size and S, the selective disadvantage are of order 1 or less. Moreover,
new mutations are almost always immediately lost to the population even when they
are selectively advantageous. The proba-
819
GENETICS IN EVOLUTION
bility of the eventual incorporation of new
mutation is only 2S, where S is the selective
advantage of that mutation. Of course, if
the same mutation occurs over and over
again it may expect to succeed. But the
time and population available are not infinite and many favorable mutations will fail.
Second, random factors may interact
with selection to produce non-selective differentiation. Wright's models of adaptive
peaks (1931) show that if genes interact
epistatically in development there may be
multiple stable outcomes to the same selective process. Which of these outcomes will
be realized by a particular population
depends upon chance fluctuations in gene
frequency. As a result two populations may
be driven apart phenotypically by natural
selection even though they are both
responding to identical selective forces.
Selective differentiation does not mean differentiation of selection.
Third, the structures of inheritance
themselves may drive genes non-selectively. The organization of genes on chromosomes means that if a locus is under
selection and allele frequencies are changing, alleles at other loci, linked to the first
but not under direct selection, will also
evolve by the phenomenon of "genetic
hitch-hiking." This linkage effect may cause
pseudo-selective changes in totally unselected genes or counter-selective changes
in weakly selected genes. The organization
of genes on chromosomes also means that
any abnormality in the process of chromosome replication or segregation will
drive gene frequencies or that chromosome. Meiotic drive, in which chromosomes that carry a particular allele or chromosome segment are preferentially
included in gametes, is a powerful mechanism for changing the frequencies of
whole suites of unselected characters.
Fourth, genes have multiple developmental effects so that selection of one effect
may cause unselected or even counterselective changes in others. Eye pigments
are probably under selection in Drosophila,
at least in part for mating success. The same
pigments are deposited in the Malpighian
tubules, but these are never observed by
any sensory apparatus. Not only genes, but
all physical events have multiple effects.
Selection has provided us with hemoglobin
to carry oxygen. Our blood is red by physical accident and some animals have green
blood if they have a copper haem pigment.
Only in a trivial sense has selection made
our blood red, an epiphenomenon of iron
porphyrin structure. In addition, of course,
there are the negative developmental correlations that arise from the pleiotropic
actions of genes and which are then
unbreakable in the absence of quite new
kinds of mutations or drastic reconstruction of development.
Thus, the findings of developmental and
population genetics have not simply confirmed the possibility of evolution of new
species by the selection of heritable variants, the core of 19th century Darwinism,
but have given rise to a vastly enriched
20th century neo-Darwinism which still
understands the origin of taxonomic diversity as being rooted in individual variability, but which greatly extends the complexity of that process.
REFERENCES
Chinnici.J. P. 1971. Modification of recombination
frequency in Drosophila I. Selection for increased
and decreased crossing over. Genetics 69:71-83.
Dobzhansky, T h . 1937. Genetics and the origin of species.
Columbia University Press, New York.
Gupta, A. P. and R. C. Lewontin. 1982. A study of
reaction norms in natural populations of Drosophila pseudoobscura. Evolution 36:934—948.
HaldaneJ.B.S. 1932. Thecausesojevolution. Harper,
London & New York.
Fisher, R. A. 1930. The genetical theory of natural selec-
tion. Clarendon Press, Oxford.
Keith, T. P., L. D. Brooks, R. C. Lewontin, J. C.
Martinez, and D. L. Rigby. 1985. Nearly identical allelic distributions of xanthine dehydrogenase in two populations of Drosophila pseudoobscura. Mol. Biol. Evol. 3:206-216.
Kreitman.M. 1983. Nucleotide polymorphism at the
alcohol dehydrogenase locus of Drosophila melanogaster. Nature 304:412-417.
Mather, K. 1953. Genetic control of stability in
development. Heredity 7:297-336.
Maynard Smith, J. and K. Sondhi. 1960. The genetics
of a pattern. Genetics 45:1039-1050.
Mayr, E. 1963. Animal species and evolution. Harvard,
Belknap Press, Cambridge.
Ramshaw, J. A., J. A. Coyne, and R. C. Lewontin.
1979. The sensitivity of gel electrophoresis as a
detector of genetic variation. Genetics 93:10191037.
820
R. C. LEWONTIN
Robertson, F. W. and E. C. R. Reeve. 1952. Studies
alleles of xanthene dehydrogenase in Drosophila
pseudoobscura. Genetics 84:609-629.
in quantitative inheritance I. The effects of selection of wing and thorax length in Drosophila me- Waddington, C. H. 1942. Canalization of developlanogaster. J. Genet. 50:414-448.
ment and the inheritance of acquired characters.
Scharloo, W. 1964. The effects of disruptive and
Nature 150:563-565.
stabilizing selection on the expression of a cubitus Waddington, C. H. 1960. Experiments on canalizing
interuptus mutant in Drosophila. Genetics 50:553selection. Genet. Res. Cambridge 1:140-150.
562.
Wright, S. 1931. Evolution in Mendelian populaSingh, R. S., R. C. Lewontin, and A. Felton. 1976.
tions. Genetics 16:97-159.
Genetic heterogeneity within electorophoretic